ISSN 1536-9323
Journal of the Association for Information Systems (2021) 22(5), 1334-1375 doi: 10.17705/1jais.00696 RESEARCH ARTICLE
The Mechanics of Enterprise Architecture Principles
Kazem Haki1, Christine Legner2
1University of St. Gallen / Geneva School of Business Administration (HES-SO), Switzerland, [email protected]
2Faculty of Business and Economics (HEC), University of Lausanne, Switzerland, [email protected]
Abstract
Inspired by the city planning metaphor, enterprise architecture (EA) has gained considerable attention from academia and industry for systematically planning an IT landscape. Since EA is a relatively young discipline, a great deal of its work focuses on architecture representations (descriptive EA) that conceptualize the different architecture layers, their components, and relationships. Beyond architecture representations, EA should comprise principles that guide architecture design and evolution toward predefined value and outcomes (prescriptive EA).
However, research on EA principles is still very limited. Notwithstanding the increasing consensus regarding the role and definition of EA principles, the limited publications neither discuss what can be considered suitable principles nor explain how they can be turned into effective means to achieve expected EA outcomes. This study seeks to strengthen the extant theoretical core of EA by investigating EA principles through a mixed methods research design comprising a literature review, an expert study, and three case studies. The first contribution of this study is that it sheds light on the ambiguous interpretation of EA principles in the extant research by ontologically distinguishing between principles and nonprinciples, as well as deriving a set of suitable EA (meta)principles. The second contribution connects the nascent academic discourse on EA principles to studies on EA value and outcomes. This study conceptualizes the “mechanics” of EA principles as a value-creation process, where EA principles shape architecture design and guide its evolution and thereby realize EA outcomes. Consequently, this study brings the underserved, prescriptive aspect of EA to the fore and helps enrich its theoretical foundations.
Keywords: Enterprise Architecture, Enterprise Architecture Principles, Enterprise Architecture Value, Mixed Methods Research
Jan vom Brocke was the accepting senior editor. This research article was submitted on January 9, 2019 and underwent two revisions.
1 Introduction
Since the 1980s, practitioners and academic scholars have propagated the notion of architecture as an approach to systematically planning and developing IT landscapes (Earl, 1993; Lederer & Sethi, 1988; Segars
& Grover, 1998; Zachman, 1987; Zachman, 1997).
The similarities between city planning and the IT domain, which both deal with complex supersystems
and require ongoing management to address various stakeholders’ constantly changing interests, have inspired the seminal publications and theoretical concepts that underpin the enterprise architecture (EA) discipline. In his pioneering work, Zachman (1987) built on architecture abstraction and proposed a framework to systematically document EA, with different representation types addressing different stakeholder concerns. In another early, seminal EA publication, Richardson et al. (1990) took a different,
yet complementary, stance, emphasizing a principles- based EA. In their view, principles reflect “the organization’s basic philosophies that guide the development of the architecture” and have a “far- reaching and significant impact on an organization because they are the most stable element of an architecture” (Richardson et al., 1990, p. 389).
Today, the architecture concept is acknowledged as playing a fundamental role in the design of an organization as a complex adaptive sociotechnical system (Haki et al., 2020; Schmidt & Buxmann, 2011) and in guiding its transformation from a current state to a future state (Lange et al., 2016). EA is seen as a coherent unity of principles, methods, and models providing a blueprint for organizations (Lankhorst, 2009, p. 3; Ross et al., 2006, p. 9). Many EA studies refer to the ISO/IEC/IEEE 42010:2011 Standard (ISO/IEC/IEEE, 2011) to characterize [enterprise]
architecture as “fundamental concepts or properties of a system in its environment embodied in its elements, relationships, and in the principles of its design and evolution.” In both definitions, EA comprises both descriptive and prescriptive aspects,1 an understanding that other architectural disciplines share and which dates back to the Roman author, architect, and civil engineer Vitruvius and his De architectura.
In EA, the descriptive aspect builds on Zachman’s tradition (Sowa & Zachman, 1992; Zachman, 1987, 1999) and is associated with the artifacts representing an organization in its as-is and to-be states. 2 Descriptive EA focuses on creating architecture representations to depict and explain an organization’s design as a sociotechnical system and in terms of its constituents, properties, and relationships. In turn, the prescriptive aspect takes Richardson et al.’s (1990) stance and emphasizes the principles governing the design and evolution of architecture. Prescriptive EA draws attention from architecture representations, in the form of artifacts, toward the architectural shape and the question of “how” organizations should be designed and built. The prescriptive aspect therefore comprises principles to guide an organization’s design and evolution to achieve predefined outcomes and a desired future state.
1 Besides descriptive and prescriptive aspects, Bean (2010) proposes a programmatic strand of EA, which concerns the design of and migration toward a target architecture. We argue that this strand is inherently descriptive by nature because it proposes describing a target state explicitly in terms of an architecture model and its components. Therefore, in line with ISO/IEC/IEEE 42010:2011, as well as other conceptualizations (e.g., Fischer et al., 2010; Hoogervorst, 2004; Winter & Aier, 2011), we distinguish between the descriptive and prescriptive aspects of EA.
2 Our understanding of descriptive EA reflects the nature of many architecture artifacts, specifically models and modeling notations that explicate how the architecture
EA is a still maturing discipline (Boh & Yellin, 2006;
Schmidt & Buxmann, 2011; Tamm et al., 2011) and has long focused on the descriptive aspect. In comparison, the number of publications related to prescriptive EA is very limited—a research gap that other studies have also noted (Greefhorst & Proper, 2011; Stelzer, 2010). By enriching the theoretical foundations of EA, this study seeks to bring the underserved, prescriptive aspect to the forefront of EA research. We posit that in dealing with hyperturbulent and dynamic environments, EA needs to be sufficiently agile to constantly adapt IT landscapes to ever-changing organizational and technological requirements (descriptive, the constantly changing aspect of architecture) (Haki & Legner, 2013a; Nan &
Tanriverdi, 2017; Tanriverdi et al., 2010). While such an adaptation process is required to survive and thrive in the environment, it also bears the risk of making the evolution of architecture inherently emergent and its outcomes inevitably unpredictable (Benbya et al., 2020;
Nan, 2011). Therefore, beyond architecture requiring a plastic core to evolve dynamically with environmental changes, it requires a set of principles as a robust core in order to purposefully guide its evolution (prescriptive, the most stable aspect of architecture). Such principles are crucial to ensure the guided, rather than entirely emergent, architecture evolution to obtain the predefined value and outcomes of EA (Haki et al., 2020). Drawing on city planning and architectural concepts, we therefore postulate that, without explicating EA principles, the knowledge inherent in EA’s as-is and to-be design cannot be shared and developed further.
Previous studies have been instrumental in creating a basic understanding by delineating the definition and formulation of EA principles, although the nascent discourse on these principles still lacks consolidation and theoretical integration: First, prior work suggests either company-specific principles, which may not be generalizable, or proposes generic principles, which are not explicitly studied in the EA context. At the same time, prior studies remain ambiguous in their interpretation of EA principles in terms of their nature and raison d’être to guide design decisions. Second, the debate on EA principles is fragmented and largely isolated from EA value and outcomes research. We therefore know little about how principles (as carriers of
should be modeled and represented. While they essentially help represent as-is and to-be designs, they do not entail prescriptive knowledge about a “good” or a “bad”
architecture design. Consequently, these artifacts cannot be considered prescriptive in the sense of prescribing how the architecture should be built. Our view therefore reflects the general knowledge base of the EA discipline, and should not be confused with the prescriptive function of EA in a specific company’s context, where the to-be EA is meant to “guide and constrain the subsequent development of business and IT solution” (c.f. Gong & Janssen, 2019).
knowledge about good design) can be used as an effective means to achieve EA value and outcomes. To address these gaps, we investigate the following research questions:
RQ1: What are suitable EA principles to guide architecture design and evolution?
(metaprinciples)
RQ2: How do EA principles contribute to achieving EA outcomes? (the mechanics of EA principles) In answering RQ1, we ontologically analyze the suggested EA principles in the literature and specifically highlight the phenomenon of nonprinciples, i.e., EA principles from academic and practitioner literature that do not conform with the basic understanding of EA principles as a high-level governance instrument.
Thereafter, based on their content similarities, we group principles into metaprinciples that provide architecture design and evolution with specific guidance. The focus on metaprinciples, instead of detailed and overlapping principles, provides us with a thorough understanding of the mechanics of EA principles by uncovering their joint contributions and complementarity. Our study thereby contributes to advancing the design knowledge inherent in EA principles and making it accessible to the EA and IS research communities.
In answering RQ2, this study conceptualizes the mechanics of EA principles by employing a value- creation approach (Schryen, 2013). Instead of pursuing an outcome-oriented (i.e., deterministic, means-end relations) approach, our study sheds light on the ways that EA principles (as means) shape architecture design and guide its evolution to create EA outcomes (as ends).
This study empirically illustrates instantiations of the implications of metaprinciples in obtaining EA outcomes, and reveals complementary relations between metaprinciples as an integral part of their mechanics.
In this study, we opted for an exploratory research design and employed a mixed methods research process.3 To reflect the research process and its steps, the remainder of this paper is structured as follows: First, we critically review the current status of the research, and outline four distinct research axes. Thereafter, we motivate and present our research design and process. In response to RQ1, we present our ontological analysis and insights from expert studies to derive a set of metaprinciples. We subsequently answer RQ2 by conceptualizing and empirically illustrating the mechanics of EA principles.
We conclude by discussing our results and providing a research outlook.
3 By following a cumulative research design, this paper complements two conference publications with preliminary research findings derived from the literature review and the expert study (Haki & Legner, 2013b; Haki & Legner, 2012).
The manuscript at hand integrates the preliminary research
2 State of the Research
2.1 Evolution of the Enterprise Architecture Discipline
EA is a still maturing discipline, mostly driven by practice and underrepresented in the top academic journals. To illustrate the evolution of the EA discipline, Table 1 provides a synthesis of influential EA publications in research and practice that initiated the main discourses of the EA discipline along with their contributions to EA foundations, value, and outcomes. This synthesis builds on some of the first publications that gave rise to the decisive and influential discourses of the EA discipline (i.e., EA frameworks, principles, maturity models, modeling, governance, success, and organizational benefits), regardless of the publication type and outlet. We consequently also acknowledge the influence of the discipline’s practitioners.
Although the twin descriptive-prescriptive aspects have been inherent in the EA concept since the earliest contributions (as two sides of the same coin), the extant literature mostly emphasizes the descriptive side. After the early descriptions of architecture artifacts (Sowa &
Zachman, 1992; Zachman, 1987), the isolated representations were later integrated into and/or complemented by EA frameworks, such as the enterprise architecture planning (EAP) framework (Spewak & Hill, 1993) and the Open Group Architecture Framework (TOGAF) (The Open Group, 2011, 2018). The EA frameworks were accompanied by a great deal of research to develop modeling techniques (Johnson et al., 2007; Jonkers et al., 2003;
Lankhorst et al., 2004), and to propose EA methods (Peristeras & Tarabanis, 2000; Wegmann, 2002) and maturity models (Ross, 2004; Ross et al., 2006;
Venkatesh et al., 2007). The key contributions of descriptive EA research are therefore the conceptualization, description, and modeling of EA layers and components, but also the development of specific EA modeling notations. Nonetheless, during the last decade, research interest has shifted from EA representation and modeling to EA value and the more holistic EA management (EAM) concept. The increasing academic discourse on EA has resulted in publications in leading journals examining EA governance (Boh & Yellin, 2006), success (Lange et al., 2016; Schmidt & Buxmann, 2011), and organizational value (Tamm et al., 2011).
findings into a comprehensive research model and extends them by analyzing the mechanics of EA principles, which are based on three case studies. Since this manuscript builds on two prior publications, their fundamental inputs are also included.
Table 1. Overview of Influential EA Publications
Reference Focus of research Contribution to EA foundations Contributions to EA value and outcomes Descriptive EA Prescriptive EA
Zachman (1987, 1999)
EA framework (practice-oriented)
An EA framework to systematically represent architecture artifacts and organize architecture models
-
-
Richardson et al.
(1990)
EA principles
-
A set of EA principles in an exemplary case company
- Open Group,
(2011) – TOGAF
EA framework (practice-oriented)
An EA framework developed by practitioners to step-by- step design, plan, implement, and govern EA
A tentative catalogue of principles
-
Lankhorst (2004) EA modeling (practice-oriented)
An EA modeling language
- -
Ross et al. (2006) EA maturity model and guidelines (practice-oriented)
Architecture maturity model, core architecture diagrams, operating models
- -
Boh & Yellin (2006)
EA governance
-
EA standards as unifying principles
Impact of governance mechanisms and EA standards on EA outcomes Schmidt &
Buxmann (2011)
EAM success factors and outcomes
- -
Success factors to attain architecture outcomes Tamm et al., (2011) EA value to
organizations - -
A set of EA benefit enablers leading to organizational benefits
Although the literature has widely discussed the descriptive aspect, the prescriptive aspect (i.e., EA principles) still remains the crux of the EA concept.
After the seminal work by Richardson et al. (1990), the academic literature has remained silent about EA principles, with the exception of Boh and Yellin’s (2006) study on EA governance. The latter presents EA standards as unifying principles that influence technical choices and decisions related to the data and the application design across projects and business units. Standards can thus be associated with prescriptive EA, even though, in Section 4.1, we provide a more fine-grained distinction between principles and standards (as a course of action and set of rules for principles). It is also noteworthy that TOGAF, the most popular EA framework, comprises a tentative catalogue of EA principles.
Overall, this synthesis of influential EA publications reflects the priorities of the EA discipline: earlier, mostly practitioner-oriented publications focused on the foundations of the EA discipline, which comprised the descriptive and, to a lesser extent, the prescriptive aspects, whereas the more recent top journal publications investigate EA value and outcomes.
2.2 EA Principles as Normative Principles
Generally speaking, principles denote “comprehensive and fundamental laws, doctrines, or assumptions”
(Merriam-Webster, 2003) and provide insights into the causes of certain effects, which are rooted in laws of nature, facts, or beliefs (see Greefhorst & Proper, 2011). Principles are fundamental concepts of any engineering discipline, such as civil, mechanical, or software engineering, that emphasize the design of artifacts. They are important instruments for explicating and sharing design knowledge and the rationale that guides design decisions.
EA principles fall into the normative principle category, as they influence an organization’s construction and architectural design. Table 2 illustrates the distinction between scientific and normative principles, which Greefhorst and Proper (2011) have discussed extensively. Scientific principles are based on laws of nature (e.g., the law of gravity) and do not change with time or distance. They are, therefore, the same today as they were millions of years ago, although their scope of applicability has
changed. While scientific principles underlie the working of human-made artifacts, normative principles represent “rules of conduct” related to artifacts’ design. These normative principles are thus based on fundamental beliefs and assumptions about how things should or ought to be, and how they should be valued in terms of good or bad. These principles restrict freedom of action normatively in order to achieve expected goals. Normative principles are derived from experience and expert knowledge. They are stable and enduring, but new practices and knowledge can change them. Compared to scientific principles that hold naturally, normative principles need enforcement procedures to be put into place.
An EA principle is “included in architecture” and represents “a declarative statement that normatively prescribes a property of the design of an artifact, which is necessary to ensure that the artifact meets its essential requirements” (Greefhorst & Proper, 2011, p.
44). In other words, each EA principle reflects specific architecting knowledge and is derived inductively from EA practice to synthesize knowledge about designing “satisficing artifacts.” Therefore, as the theoretical core of a design inquiry, EA principles are design principles that capture, synthesize, and share essential architectural design knowledge (Chandra et al., 2015; Gregor & Jones, 2007; vom Brocke et al., 2020), and allow for the projection of the design knowledge beyond instantiations that are applicable in a limited use context (Baskerville & Pries-Heje, 2019).
2.3 The Nascent Academic Discourse on EA Principles
Since the 2010s, several authors have begun to acknowledge EA principles as the cornerstone and an integral part of EA but have also pointed out that EA principles remain an underexplored aspect of the EA concept (Aier et al., 2011; Op’t Land & Proper, 2007;
Proper & Greefhorst, 2011; Stelzer, 2010; Winter &
Aier, 2011). This debate has resulted in a number of conference publications but has not yet been incorporated into the existing body of academic knowledge to enrich EA foundations (see Table 1).
Consequently, research on EA principles is not easily accessible to the broader IS research community. To assess this nascent discourse on EA principles, we conducted a literature review and classified prior publications based on the primary IS research objectives and theory types that Gregor (2006) suggested. While different types of theory are closely interrelated, Gregor’s (2006) taxonomy systematically
4 The notion of EA principles guiding the design and evolution of architecture should not be confused with principles, enablers, or factors for successfully deploying EA as a function
distinguishes theory types with regard to their distinct goals and attributes. Building our investigation of EA principles on Gregor (2006) thereby allowed us to identify the theoretical contributions that EA principles research should provide and assess how the relevant literature has addressed them. This analysis resulted in the identification of four research axes in EA principles research (see Table 3).
2.3.1 Nature (What Are EA Principles?) This research axis investigates the what, in other words, the definition and characteristics of the phenomena of interest, resulting in theory type I (theory for analyzing) in Gregor’s taxonomy, which is the most basic theory type. In this research axis, theoretical contributions lay the groundwork for other theory types by providing basic definitions, classification schema, taxonomies, or typologies. On assessing prior studies (see Table 3), we found that they predominantly focus on this axis by: (1) suggesting an exhaustive and comprehensive definition of EA principles and shedding light on their role, (2) discussing the formulation and statement of EA principles as a set of constraints regarding the syntax and semantics of the documentation of EA principles, and (3) categorizing EA principles into different areas and scopes. The most important contribution of the first research axis is a shared understanding of the role, definition, and documentation of EA principles as follows.
EA principles are used to govern architecture design and evolution, and to limit design space and guide architecture design decisions (Op’t Land & Proper, 2007; Stelzer, 2010; Van Bommel et al., 2007; Van Bommel et al., 2006). These principles can be attributed to different architectural layers, should be based on business and IT strategies, and refer to an organization’s construction. Since architecture is about the aligned (re)design of an organization’s technological (i.e., IT components and their relations) and organizational (e.g., business processes) constituents, EA principles refer to principles that guide such essential design decisions in order to achieve predefined outcomes.4 For complete and exhaustive documentation, each EA principle should be described in a principle statement, along with a rationale that explains why that principle is helpful to achieve predetermined outcomes, and the implications that describe how to implement this principle. Finally, metrics should be identified for each principle to measure its fulfillment (Aier et al., 2011;
Fischer et al., 2010; Lindström, 2006; Richardson et al., 1990; Van Bommel et al., 2006).
(e.g., implementing EA frameworks or modeling techniques in specific companies’ context) to achieve a high-quality EA function (Niemi & Pekkola, 2013, 2016).
Table 2. Normative vs. Scientific Principles
Normative principles Scientific principles Nature of principles Declarative statements that define the
artifact properties (“what should be”) Scientific rules that govern the working artifact (“what is”)
Causes of principles Facts and fundamental beliefs Laws of nature Formulation/derivation of
principles
Inductive (based on expert knowledge and experience)
Deductive (derived from laws of nature)
Table 3. Current State of Literature Related to EA Principles Research axis Theory type (Gregor, 2006) Current state and research directions Nature: What are EA
principles?
Theories for analyzing (type I) Consensus on EA principles definition and documentation:
1. The definition and role of EA principles (Aier et al., 2011;
Armour et al., 1999; Chen et al., 2008; Fischer et al., 2010;
Greefhorst & Proper, 2011; Hoogervorst, 2004; Proper &
Greefhorst, 2010, 2011; Stelzer, 2010; Van Bommel et al., 2007;
Winter & Aier, 2011; Sandkuhl et al., 2015; Greefhorst et al., 2013);
2. The formulation and documentation of EA principles (Lindström, 2006; Richardson et al., 1990; Van Bommel et al., 2007; Van Bommel et al., 2006; Greefhorst et al., 2013; Marosin et al., 2016);
3. Categorizing EA principles (Lindström, 2006; Op’t Land &
Proper, 2007; Richardson et al., 1990; Winter & Fischer, 2007).
Practices: How does one design, implement, and manage EA principles?
Theories for design and action (type V)
Tentative or implicit processes for principle extraction and management, as well as some sample principles:
1. The extraction process of EA principles (Aier et al., 2011; Fischer et al., 2010; Greefhorst & Proper, 2011; Winter & Aier, 2011);
2. The life cycle management of EA principles (Greefhorst &
Proper, 2011; Op’t Land & Proper, 2007; Van Bommel et al., 2007; Winter & Aier, 2011; Uludağ et al., 2019; Sandkuhl et al., 2015);
3. Sample EA principles (Janssen & Kuk, 2006; Lindström, 2006;
Nightingale, 2009; Richardson et al., 1990; Wilkinson, 2006).
Adoption: Why, how, and to what extent are EA principles adopted?
Theories for explaining (type II)
First empirical insights, but no general theories on the adoption of EA principles:
1. Moderating role of organizational culture (Aier, 2014);
2. Main challenges in establishing EA principles (Uludağ et al., 2019).
Impact: What are the impacts of EA principles?
Theories for explaining and predicting (type IV)
First attempts to examine EA principles’ impact:
1. The impact of EA principles on EA consistency and EA utility (Aier, 2014);
2. The impact of EA principles on managing IT investments and sustainable business-IT alignment (Pessi et al., 2014; Pessi et al., 2011).
2.3.2 Practices (How Does One Design, Implement, and Manage EA Principles?)
This axis specifies guidelines on how organizations should develop, deploy, and manage EA principles.
Gregor (2006) classifies this research axis as theory type V (theory for design and action) and associates it with a constructivist type of research or design science.
Prior contributions related to this research axis can be categorized into three different areas: (1) the generic
process of determining or extracting principles, (2) managing the lifecycle of principles in order to turn these principles into effective means to guide EA design and evolution, and (3) suggesting either a set of company-specific EA principles or principles that are not explicitly studied in the EA context.
2.3.3 Adoption (Why, How, and to What Extent Are EA Principles Adopted?) This research axis comprises approaches to analyze the adoption and diffusion of EA principles in different
organizational contexts. Studies in this research axis ultimately lead to insights into adoption patterns and the factors that underlie or explain the organizational adoption of EA principles, generating theory type II (theory for explaining) in Gregor’s taxonomy. Prior research has not yet adequately embraced this research axis. Exceptions are Aier (2014), who illustrates how organizational culture moderates the organizational adoption of EA principles, as well as Uludağ et al.
(2019), who investigate how EA principles are established and discuss the associated challenges.
2.3.4 Impact (What Are the Impacts of EA Principles?)
This research axis considers the theoretical constructs and relationships between them in order to explain and predict their impacts. The research in this axis generally results in theory type IV (theory for explaining and predicting) in Gregor’s taxonomy. In EA, the measuring of impacts and organizational outcomes are very important, but we found only a few studies (Aier, 2014; Pessi et al., 2014; Pessi et al., 2011) related to the impacts of EA principles. Aier (2014) suggests that the grounding, management, and guidance of EA principles improve the consistency and utility of EA and suggests that EA principles have an indirect effect on EA outcomes. Pessi et al. (2014, 2011) argue that the choice of EA principles impacts both (1) the ability to achieve and maintain a sustainable business-IT alignment in a dynamic business context, and (2) the responsibility for IT investments and the coordination of such investments with business changes.
2.4 Research Gap
EA principles are normative principles that should be used in the constant examination and reevaluation of a proposed IT target plan (Richardson et al., 1990;
Stelzer, 2010) toward the expected value and outcomes. Although there is increasing consensus regarding the nature of EA principles and suggestions for their definition and documentation, knowledge related to the principles governing the design and evolution of architecture is fragmented and not systematically accumulated. Beyond EA design knowledge, we know little about what constitutes a suitable principle in the EA context. Prior work suggests either company-specific principles, which might not be generalizable, or proposes generic principles, which are not explicitly studied in the EA context. In addition, studies on EA principles are largely isolated and not connected to the discourse on EA value and outcomes. The extant literature thereby disregards how EA principles shape the architecture design and evolution, and how they contribute to achieving the expected EA value and outcomes.
To strengthen the theoretical core of the EA discipline, this study consolidates the existing body of knowledge on EA principles and addresses three important research gaps outlined by existing studies:
• Proposing a set of suitable principles with the potential to act as effective means to guide architecture design and evolution (Radeke, 2011;
Stelzer, 2010);
• Studying the roles and usefulness of principles in EA endeavors (Greefhorst & Proper, 2011); and
• Investigating the relationship between deploying EA principles and achieving architecture value and outcomes (Fischer et al., 2010; Stelzer, 2010;
Winter & Aier, 2011).
3 Research Method and Approach
To synthesize suitable EA principles (RQ1) and investigate their mechanics (RQ2), we opted for a mixed methods exploratory research design (Creswell & Clark, 2011; Tashakkori & Teddlie, 2010), and combined a literature review, an expert study, and case studies (see Table 4). Mixing methods can lead to new insights and modes of analysis that are unlikely to occur if only one method is used (Kaplan & Duchon, 1988; Venkatesh, et al., 2013). Following its reference disciplines, the use of mixed methods research is gaining momentum in IS (Venkatesh et al., 2013; Venkatesh et al., 2016) and has already been employed in investigating various IS phenomena (e.g., Ågerfalk & Fitzgerald, 2008; Cyr, et al., 2009; O’Leary et al., 2014; Turel & Bart, 2014). In our study, this approach not only helped us critically assess the current body of research, but also assisted us in matching the current literature with insights from subject matter experts and in-depth empirical investigation.
Our research process is organized into three steps, with each step informing the theory building. Appendix A introduces our study’s key terms along with their investigation in each step of the study and findings. In Step 1, we reviewed the extant literature to extract assumptions about the nature and role of EA principles, collect and critically assess the proposed EA principles, and develop the initial conceptualization on the mechanics of EA principles.
For this purpose, we carried out a systematic literature review of scientific journal and conference publications based on the guidelines provided by Webster and Watson (2002) and vom Brocke et al.
(2015). A set of key terms (i.e., “principle” and
“architecture principle”) was used to identify the related publications. In our search, we included articles in an EA[M] context and excluded articles addressing principles in other fields, such as in modeling (Balabko
& Wegmann, 2006; Brown, 2004). Owing to the paucity of publications on EA principles, we did not apply any publication date limitation.
Table 4. Research Process
Steps and contribution Tasks Outcomes
Literature review (S1)
→ Ontological analysis to distinguish between principles and nonprinciples
Extract insights from literature (S1.T1)
• A set of assumptions about the definition, roles, and usefulness of EA principles
• An initial conceptualization of the mechanics of EA principles Collect EA principles from the
literature (S1.T2)
152 nonunique principles and their statement, implications, and rationale
Conduct ontological analysis to critically assess the proposed principles in the literature (S1.T3)
• Ontological analysis of the collected principles (S1.T2) in comparison with their basic definition (S1.T1)
• Distinguishing between principles and nonprinciples Consolidate the remaining
principles into unique principles (S1.T4)
A set of 45 unique principles
Group principles into metaprinciples based on their shared implications and rationales (S1.T5)
A consolidated list of 45 unique principles, classified into nine EA metaprinciples
Expert study (S2)
→ Experts’ opinions on EA principles in practice
Expert sampling (S2.T1) A list of experts from the Open Group Conference and EA expert communities
Conduct expert interviews (S2.T2)
• Refined and enhanced assumptions about EA principles based on S1.T1
• Refined and enhanced EA metaprinciples based on S1.T5 Conduct a semistructured
survey (S2.T3)
• Experts’ feedback on the assumptions about EA principles (from S1.T1 and S2.T2) and a set of new assumptions resulting from the open-ended questions
• A set of metaprinciples based on S1.T5 (experts deemed eight of the nine metaprinciples as practically relevant)
Case studies (S3)
→ Conceptualization and empirical illustration of the mechanics of EA principles
Case sampling (S3.T1) • Selection of three companies from different industries employing different principles
Data collection (S3.T2) Three comprehensive case write-ups
Within-case analysis (S3.T3) • Coded case write-ups based on a predefined coding scheme. The latter is developed based on the initial conceptualization of the mechanics of EA principles (S1.T1), concluded metaprinciples (from S2.T3), and EA outcomes
• EA principles, their implications, and their impacts on EA outcomes for each case as a stand-alone entity
Cross-case analysis (S3.T4) • Commonalities and differences between the employed EA principles, their implications, and their impact on EA outcomes We identified the related articles by scanning scientific
databases, namely AIS electronic library, ACM Digital Library, DBPL, EBSCOhost, IEEE Xplore Digital Library, Science Direct, Web of Science, and SpringerLink to cover a wide range of outlets since EA scholars publish in various communities. The first step of our literature review resulted in 32 articles on EA principles. We then coded and analyzed the identified articles based on Gregor’s (2006) taxonomy of theory types (see Section 2.3). In the subsequent step, we analyzed the proposed EA principles in existing research. Our primary source for identifying EA principles was peer-reviewed EA-related publications (Dietz & Hoogervorst, 2012; Janssen & Kuk, 2006;
Lindström, 2006; Richardson et al., 1990; Wilkinson, 2006). Since certain practitioner publications provide comprehensive collections of principles, we decided to include the catalog of principles that the Open
Group (2011) provides as the most important professional resource for EA experts, and the ones that Greefhorst and Proper (2011) propose in the only published book on EA principles. This effort resulted in 152 nonunique principles from the aforementioned sources. Table 5 provides an overview of sources and their proposed principles.
In the next step, we coded the identified 152 nonunique principles based on their statements, implications, and rationales. The results formed the basis of an ontological analysis to distinguish between principles and nonprinciples; we excluded the latter from further investigation. Thereafter, we consolidated similar principles and classified the remaining 45 unique principles during the course of several rounds and synthesized them into nine metaprinciples (i.e., principles that share common implications and rationale).
Table 5. Overview of the Literature on the Proposed Principles
Reference Methodology Suggested principles Documentation
Richardson et al., (1990)
Case study: Texaco and Star Enterprise
18 principles in four architectural layers:
organization, application, data, and infrastructure
Statement, rationale, and implications per principle Lindström (2006) Case study: Vattenfall 35 principles classified into governance,
outsourcing, risk management and security, system management, environment,
standardization, application, and infrastructure categories
Only list of principles
Greefhorst and Proper (2011)
Experience-based A catalogue containing 59 principles covering different architectural layers
Type of information, quality attributes, rationale, and implications per principle Open Group (2011) Experience-based 21 principles in four architectural layers:
business, data, application, and technology
Statement, rationale, and implications per principle Wilkinson (2006) Conceptual insights Modularity, simplification, integration, and
standardization (4 principles) as the main principles for adaptive EA
General description per principle
Janssen and Kuk (2006)
Insights from 11 e- government projects
8 principles from a complex adaptive system perspective
General description per principle
Dietz and
Hoogervorst (2012)
Conceptual insights 7 principles for dealing with enterprise transformation
General description per principle
Step 2 comprised an exploratory study to collect experts’ judgments on the metaprinciples resulting from Step 1 and refine the assumptions about the mechanics of EA principles. The expert study’s main objective was to complement the literature review (Step 1) and to inform theory building regarding the mechanics of EA principles (Step 3).
First, we organized exploratory interviews with two experienced enterprise architects in the banking and insurance industries. Each interview lasted two hours on average and resulted in complementary assumptions and principles based on the interviewees’
experience and observations. Second, we conducted a questionnaire-based exploratory survey because this is the most effective way to rigorously collect opinions and to ask experts to grade a variety of assessment items about a topic (Pinsonneault & Kraemer, 1993).
We prepared a questionnaire (see Appendix C) containing scale-response (on a 5-point Likert scale) and open-ended questions to collect expert feedback on our assumptions and metaprinciples. We identified experienced practitioners with a strong background and with demonstrated field expertise in developing and deploying EA principles from those attending the Open Group Conference, one of the most influential EA conferences, and from EA expert communities (reached through LinkedIn’s professional database).
The sample covered 26 experts with an average of 10 years’ experience in EA as (chief) enterprise/IT architects and representing different sectors and company sizes. All except one used EA principles, either in their affiliated companies or for their clients.
The sectors are consultancy (10), banking, insurance, government (3 each), health (2), aerospace, defense,
telecommunications, retail, and transportation (1 each). The nonconsultancy companies covered seven large ( > 5,000), four medium-sized to large (1,000 to 5,000), four medium-sized (100 to 1,000), and one small ( < 100) organization.
There are two noteworthy points about the expert study in Step 2. First, a small sample size is a typical characteristic of expert surveys (Christopoulos, 2009;
Hakim, 1987), since the population comprises particular, rarely available individuals with a specific expertise. This is even more decisive in the field of EA principles, which is considered an uncharted EA territory (Greefhorst & Proper, 2011). Second, the expert survey in Step 2 did not follow the purpose of conventional surveys (Hawlitschek et al., 2016;
Otjacques et al., 2007; Pinsonneault & Kraemer, 1993) because of its exploratory nature and complementary usage. It was employed as a complementary step to the literature review (Pinsonneault & Kraemer, 1993) for the purpose of systematically collecting experts’
judgments and opinions on the types of and reasons for employing EA principles (see Appendix D).
At the end of Step 2, we had EA metaprinciples and derived an initial conceptualization of their mechanics.
However, we still lacked insights into how these principles restrict architecture design freedom and how they contribute to achieving the expected EA outcomes. In Step 3, we therefore conducted multiple- case studies to empirically study how the most prominent EA principles impact architecture design and evolution and create value to eventually theorize the mechanics of EA principles. Case studies provide an understanding of the dynamics present within and between single settings (Benbasat et al., 1987;
Eisenhardt, 1989), which serve the purpose of our study on theorizing how EA principles contribute to achieving desirable EA outcomes. Specifically, the case studies allowed us to observe how principles shape architecture design, guide architecture evolution, and contribute to achieving EA outcomes.
When conducting the case studies, we followed the guidelines and steps set out by Yin (2003). Since EA is expected to be most useful for large organizations (Boh & Yellin, 2006; Schmidt & Buxmann, 2011;
Tamm et al., 2011), we selected three firms with more than 5,000 employees and applied additional selection criteria to follow a replication logic (Yin, 2003) to ensure cross-case diversity and generalizability of findings (Dubé & Paré, 2003): (1) The selected cases had a long history in architecture initiatives and explicitly used principles to guide their business and IT design decisions. This allowed us to observe how the principles shape the design and guide the evolution of architecture over time. (2) The selected cases covered highly ranked metaprinciples identified in our expert study in the previous step. (3) They also exploited some common metaprinciples, which allowed us to identify the commonalities and the differences in the implications of the same metaprinciple across the cases. Table 6 summarizes the main characteristics of the three cases based on the abovementioned criteria and the employed data collection sources.
Primary data were collected by means of semistructured interviews with several key informants (Yin, 2003). We conducted between three and six interviews (12 in total) in each company, ensuring that we interviewed the key informants in the architecture roles with significant expertise in business processes/business domains, applications and technology, as well as with oversight on the enterprise architecture (chief architects or CIO). Each interview was conducted by two researchers and lasted up to 150 minutes. The interview questions focused on the case company’s turning points in the architecture design, the underlying principles that guided and led to such architecture designs, and the obtained value and outcomes. We also requested the interviewees to provide us with complementary documents that we could use as secondary data. Each interview was recorded and transcribed. Transcripts and collected documents were used to prepare comprehensive case write-ups (20 to 25 pages each), and to summarize the empirical data into a consistent whole. Consequently, instead of transcribing each interview as a separate document, we reconciled the interview material with the secondary data and undertook one comprehensive case write-up per case company. The reconcilement involved different (internal and external) perspectives and several rounds with regard to each transcript to ensure an intersubjective case description and a high degree of validity. Finally, we provided the interviewees
with the comprehensive case write-ups and collected their signatures as a proxy for our full understanding of the relevant case.
Following the steps set out by Eisenhardt (1989), the data analysis was structured into early analysis and coding, within-case analysis, and cross-case analysis.
In order to familiarize ourselves with each case as a stand-alone entity, we coded each case and extracted EA principles, their impact on architecture design decisions, and the realized EA outcomes at each stage of the architecture evolution over time. We relied on a coding scheme (Miles & Hubermann, 1994, p. 55) that reflects the metaprinciples, along with their expected implications and rationale resulting from Steps 1 and 2, and the extracted EA value and outcomes provided by the extant EA literature (see Section 5.1 and Appendix E). Consequently, for each case company as a stand-alone entity, our coding endeavor identified several architecture episodes over time. Further, in each episode, we identified architecture turning points, their guiding principles, and the obtained value and outcomes. Once we had the principles, implications, and outcomes of each case at the different stages of their architecture evolution, we undertook cross-case analysis, which involved a detailed search for the commonalities and differences between the cases. The latter aimed to make cross-case inquiries, such as identifying the common implications and outcomes of different metaprinciples in different cases, the diverging implications and outcomes of the same metaprinciples in different cases, and identifying the commonalities of the same metaprinciples in different cases. Finally, we synthesized the insights from expert judgment (Step 2) and case studies (Step 3) to theorize suitable EA principles and their mechanics.
4 EA Principles
4.1 Ontological Analysis
As outlined earlier, the existing knowledge about EA principles is fragmented and has not been systematically accumulated. To consolidate academic and practitioner knowledge and derive a set of suitable EA principles, we collected 152 EA principles from the literature (see Table 5) and coded them based on their statements, implications, and rationales. We found that the EA literature is still ambiguous regarding the interpretation of EA principles and their related terms, even though several researchers (Lindström, 2006; Richardson et al., 1990;
Van Bommel et al., 2007; Van Bommel et al., 2006) have sought to clarify the notion of EA principles. We concluded that to resolve this terminological confusion, which Greefhorst and Proper (2011) have also observed and criticized, an ontological analysis is needed at the outset, as this would clarify EA principles and the vocabulary of their related terms.
Table 6. Overview of Cases and Data Collection Sources Case Size (at the
time of study)
Industry EA initiatives
EA principles Interviewees Secondary data sources A More than €150
billion in revenue, more than 500,000 employees
Auto- motive
Several architecture initiatives, earliest in 2000; major extension in 2009
Standardization from 2004 onwards;
complemented by reusability from 2007 onwards
• Chief architect CTO group, leader of technical domain architects
• Enterprise architect with a focus on application management and methods
• Enterprise architect with a business focus on SOA and modularity
• Basic organizational data and charts
• A wide range of external publications (scientific articles, case studies, and magazine articles)
• Governance reports
• EA overview presentations
• Additional information through a parallel long- term study
B Circa US$100 billion in revenue, more than 300,000 employees
Food A global program in 2000 imposing significant architecture changes
First attempted IT
standardization in 1995;
relaunch of standardization and introduction of integration at both business and IT levels from 2000 onwards
• Global CIO
• Head of business process management for one of business domains, former CFO
• Management role for technical applications
• Enterprise architect with a focus on technical perspective
• Enterprise architect with a focus on integrating overarching technical platforms
• Enterprise architect with a focus on methods
• Basic organizational data
• Global IT project progress reports and presentations
• A wide range of external publications (scientific articles, case studies, and magazine articles)
• A supervised master’s thesis in the case company
C More than €30 billion in revenue, around 100,000 employees
Bank First architecture initiatives started in early 2000s, relaunch in 2004
Reusability from 2005;
integration from 2010 onwards
• Chief architect, leader of domain architects as head of architecture
• Domain architect (business focus)
• Technical architect with a focus on SOA
• Basic organizational data
• EA overview presentations
• Domain architects’
presentations
• A wide range of external publications (scientific articles, case studies, and magazine articles)
Table 7. Three Nonprinciple Types Derived from Ontological Analysis Type of
nonprinciple
Reasons for being a nonprinciple Sample EA outcomes … are associated with the rationale of
principles, but do not limit the design space or guide the design decisions.
• Most effective use of IT as a strategic tool (Richardson et al., 1990)
• Develop competencies (Janssen & Kuk, 2006)
• Maximize benefits to the organization (The Open Group, 2011)
EA practices … describe organizational procedures that can be considered as best practices and success factors in adopting EA, but do not provide guidance or contribute to design decisions.
• IS planning as an integral part of business planning (Lindström, 2006; Richardson et al., 1990)
• Cost of IT/IS as part of a decision for M&A (Lindström, 2006)
• Primacy of principles (Dietz & Hoogervorst, 2012; The Open Group, 2011)
Low-level governance means
… concern specific guidelines for specific usages, while EA principles are pervasive by nature and concern high-level design decision points.
• Access rights must be granted at the lowest level necessary for performing the required operation (Greefhorst &
Proper, 2011)
• Using formal planning and software engineering methodologies (Richardson et al., 1990)
In our ontological analysis, we assessed the proposed EA principles against (1) the basic definition of EA principles to govern architecture design and evolution, as well as to limit the design space and guide architecture design decisions (Op’t Land & Proper, 2007; Stelzer, 2010; Van Bommel et al., 2007; Van Bommel et al., 2006), and (2) the role of EA principles as a high-level EA governance instrument (Aziz et al., 2005; Janssen & Kuk, 2006; Lindström, 2006;
Wilkinson, 2006). If the proposed EA principle did not conform with (1) and (2), we considered it a nonprinciple and excluded it from our list.
The ontological analysis allowed us to identify three nonprinciple types, representing frequent misinterpretations of an EA principle in the literature (see Table 7 for the reasons for classifying them as nonprinciples, and examples). The first category of nonprinciples comprises EA outcomes that are doubtfully formulated as principles—it confuses principles as means to improve EA capabilities (Abraham et al., 2012) with the outcomes expected from the EA. The second nonprinciple type comprises EA practices rather than EA principles. These nonprinciples describe how EA can be managed effectively (Kaisler et al., 2005; Lucke et al., 2010). Since they do not provide guidance for design decisions, they do not qualify as EA principles. The third nonprinciple type is misinterpretations of principles from the EA governance spectrum. In the hierarchy of governance means, EA principles are a high-level EA governance instrument (Aziz et al., 2005; Janssen & Kuk, 2006; Lindström, 2006; Wilkinson, 2006) that guides every design decision toward the overarching architecture. As high- level EA governance means, principles are thus pervasive by nature and should be clearly distinguished from low-level governance means such as standards (a set of rules and course of action for principles) and guidelines (methodologies in implementation) (Korhonen et al., 2009).
4.2 Metaprinciples
The ontological analysis led us to exclude nonprinciples from the initial set of 152 principles, and to combine identical principles that were proposed in different references. This resulted in 45 unique principles.
However, these 45 unique principles have different levels of granularity and partly overlap in their implications and rationales. We therefore decided to group them into nine metaprinciples, i.e., groups of principles that share common implications and rationales. This classification of principles into metaprinciples helped us concentrate on their joint implications and rationales, which is in line with the study’s goal to explain the mechanics of EA principles.
Figure 1 summarizes our step-wise investigation of principles. In addition, Table 8 presents each
metaprinciple’s constitutive principles in the extant EA literature. Moreover, we realized that the derived metaprinciples are not necessarily new in the IS literature, even though principles are considered an underexplored topic in EA. We therefore took the existing IS literature into account when synthesizing the general characteristics of the metaprinciples.
Integration: Enterprise integration comprises a set of methods, models, and tools to analyze, design, and maintain an enterprise in an integrated state (Panetto &
Molina, 2008). Companies can realize integration through, for instance, APIs and enterprise service buses that allow one application to access others’
functionalities, or through enterprise portals providing a single point of access for all applications and possibilities of information exchange along a value network (Greefhorst & Proper, 2011; Lindström, 2006).
Data consistency: Data consistency refers to the degree to which shared data definitions and consistency in stored data have been established across an organization.
It also expresses the degree to which a dataset satisfies a set of integrity constraints (Akoka et al., 2007) so that an integrated system does not lose significant functionality if the flow of services is interrupted (Panetto & Molina, 2008). By emphasizing a shared vocabulary and shared data definitions, EA principles related to data consistency seek to ensure that data are captured once, are consistent through and across all channels, are provided by the source, and support business continuity in the case of interruptions (Greefhorst & Proper, 2011;
The Open Group, 2011). Accordingly, data consistency is necessary to support system integration and denotes a complementary aspect of integrated systems (Klischewski, 2004; Panetto & Molina, 2008).
Standardization: Standardization refers to the development of company-wide standards to enable interaction between an organization’s constituent sociotechnical components (Weitzel et al., 2006).
Standardization-related EA principles recommend standardizing architectural components on different architectural layers, i.e., business processes, applications, data, and infrastructure, in order to reduce variations in all the layers and to master organizational complexity (Greefhorst & Proper, 2011; Janssen & Kuk, 2006;
Lindström, 2006; The Open Group, 2011). The adoption and realization of this metaprinciple has a higher initial cost for large organizations, owing to their size and the heterogeneity of legacy systems, but may result in considerable cost reductions in the long run (Markus et al., 2006).
Compliance: Standardization requires compliance with company-wide standards and with standards in the company’s (micro/macro)environment. This emphasizes not only the development of standards, but also their use and actual deployment (Weitzel et al., 2006).
Figure 1. The Investigation of Suitable EA Principles (RQ1)
EA principles propagate adherence to open standards (Greefhorst & Proper, 2011; Janssen & Kuk, 2006;
Lindström, 2006; The Open Group, 2011), standards created in international and/or national institutional regulatory contexts (Lindström, 2006; Lyytinen &
King, 2006; The Open Group, 2011), and company- level documents—for instance, enterprise IT architecture, and overall corporate security models (Lindström, 2006).
Reusability: Reusability-related EA principles prefer the development of applications used across the organization over the development of redundant applications (Lindström, 2006; The Open Group, 2011). Reusability is an important approach in architecture that leads to the utilization of well- established modules, which in turn improves productivity (reducing the time required to design, develop, and test), maintainability, quality, and portability (Apte et al., 1990).
Modularity: Modularity comprises a set of principles for dealing with organizations and the increased complexity of modern technologies in order to survive in a rapidly changing environment (Langlois, 2002;
Wiederhold, 1992). By shaping a system as a complex of loosely coupled components or subsystems (Weick, 1976), modular architecture allows components to be removed, replaced, and reconfigured more dynamically (Benbya & McKelvey, 2006). Proposed EA principles for modularity provide guidance for designing modular business architecture (Greefhorst &
Proper, 2011; Wilkinson, 2006), modular application architecture (Greefhorst & Proper, 2011; Janssen &
Kuk, 2006; Wilkinson, 2006), as well as multi-tier or
independent architectural layers (Greefhorst & Proper, 2011; Lindström, 2006; Richardson et al., 1990;
Wilkinson, 2006) to leverage reusability.
Usability: Usability (ease-of-use) refers to the degree to which users can associate a system’s use requirements with their existing knowledge of other systems and perceive a system’s use free of effort (Davis, 1989; Murray & Häubl, 2011). Usability has been frequently proposed as an EA principle (Greefhorst & Proper, 2011; Lindström, 2006;
Richardson et al., 1990; The Open Group, 2011) in order to achieve a shared look and feel and to support ergonomic requirements. Nevertheless, a strong emphasis on usability, particularly at the cost of functionality, is not advisable (Adams et al., 1992).
Portability: Portability-related principles foster a system’s ability to run in different computing environments (Richardson et al., 1990). This ability leads to flexibility in hardware and vendor selection. It thereby lowers the costs and facilitates migration to new technologies (Richardson et al., 1990).
Portability-related principles in EA also emphasize technology independence (Dietz & Hoogervorst, 2012;
Richardson et al., 1990).
Centralization: This EA metaprinciple concerns the centralization of application components as well as the centralization of the application development and implementation efforts within an organization (Ein- Dor & Segev, 1982; Greefhorst & Proper, 2011;
Lindström, 2006). Some scholars question the feasibility of centralization, pointing out the high associated costs (e.g., Langlois, 2002).
